
\begin{figure}[H]
\centering
\caption{RD First Stages in 2020 Platform Popup Intervention}
\label{fig:firstsagesRD2020}
%[1,1]
\makebox[\textwidth][c]{%
	\begin{subfigure}[t]{\sizeSubfig}
		\centering
		\caption{\centering Platform Pop-Up}
		%\label{ }
		\includegraphics[width=\widthPng]{figuresCL/firstStages/2020_P_fs_totalPopup_allwidth.png}
	\end{subfigure}%
%[1,2]
	\begin{subfigure}[t]{\sizeSubfig}
		\centering
		\caption{\centering SMS Warning (Pre-WhatsApp RCT)}
		%\label{ }
		\includegraphics[width=\widthPng]{figuresCL/firstStages/2020_P_fs_SMSPlain_allwidth.png}
	\end{subfigure}
	}
	\vspace{\vgap}

%[2,1]
\makebox[\textwidth][c]{%
	\begin{subfigure}[t]{\sizeSubfig}
		\centering
		\caption{\centering WhatsApp RCT Intervention}
		%\label{ }
		\includegraphics[width=\widthPng]{figuresCL/firstStages/2020_P_fs_whatsapp_allwidth.png}
	\end{subfigure}%
%[2,2]
	\begin{subfigure}[t]{\sizeSubfig}
		\centering
		\caption{\centering SMS Warning (Post WhatsApp RCT)}
		%\label{ }
		\includegraphics[width=\widthPng]{figuresCL/firstStages/2020_P_fs_SMSImage_all.png}
	\end{subfigure}
	}
\floatfoot{\footnotesize\textit{Notes.} Figure describes the receipt of various warnings interventions by risk prediction at the time of the on-platform pop-up intervention in 2020. For all graphs, risk predictions are as computed at the time of the initial platform pop-up. Panel A: count of platform pop-ups. Panel B: count of pre-WhatsApp reminder SMS interventions. Panel C:  share of individuals reading the randomized WhatsApp intervention. Panel D: share of individuals receiving follow-up SMS, after WhatsApp. Points are centered means of 50 quantile-spaced bins of the support of the predicted placement risk $\in [0.02 ; 0.98]$. Solid line shows the quadratic fit. Reported coefficients and standard errors are from local linear specifications using $+-0.1$ bandwidth.  See section \ref{sec:intdes} for intervention details.   }
\end{figure}
